Current established, evidence-based treatments for attention-deficit/hyperactivity disorder (ADHD) are incompletely effective and not universally acceptable, and appear to wane in effect over time despite significant immediate benefit. E.g., FDA-approved medication, which shows large acute benefit, leaves a third of children only partially treated even when combined with behavioral treatment, and has not been demonstrated effective beyond 2 years. Additional treatments are needed that are effective with persisting benefit, preferably related to a biomarker predicting treatment response. A good candidate is electroencephalographic (EEG) biofeedback, called neurofeedback (NF). It is based on 1) observations that patients with ADHD often have excessive theta band (4-8 Hz) quantitative EEG power, low beta band (13-21 Hz) power, and excessive theta- beta ratio (TBR), and 2) theoretical application of operant conditioning to correct this EEG imbalance. Meta- analysis of 6 randomized clinical trials found a large benefit for inattentive symptoms and medium benefit for hyperactive-impulsive symptoms. Unfortunately, none of these were blinded. Three of 4 small blinded studies found no advantage for NF over sham, but used suboptimal NF, leaving the situation inconclusive. Because of the expense and time required by NF, there is a public health need to determine whether it has a specific effect beyond the obvious nonspecific benefit of doing a focused activity several times a week with a friendly, encouraging adult who reinforces for attending to the task. Experts in NF, ADHD, clinical trials, statistics, and data management have joined to design a double-blind sham-controlled randomized clinical trial to answer several pressing scientific and clinical questions in a way that will be credible to all. At each of 2 sites (1 university & 1 NF clinic) 70 children (total N=140) age 7 through 10 with rigorously diagnosed moderate to severe ADHD and TBR>5 will be randomized in a 3:2 ratio to active TBR downtraining by NF vs. a sham training of equal duration, intensity, and appearance. To keep both participants and study staff blind, the sham will utilize pre-recorded EEGs with the participant's artifacts superimposed. The sham will be programmed into the equipment via internet by an off-site statistician-guided person who has no contact with participants. Treatment fidelity will be trained and monitored by 2 acknowledged NF leaders in a manner that protects blinding. Multi-domain assessments at baseline, mid-treatment, treatment end, and follow-ups at 6 months, 1 year, and 2 years will include parent and teacher ratings of symptoms & impairment, neuropsychological tests, clinician ratings, and quantitative EEG as well as tests of blinding and of sham inertness. Hypotheses include that NF will improve parent- and teacher-rated inattentive symptoms (primary outcome) and other outcomes more than sham,that benefit will persist for 2 years after training, that initial TBR will moderate treatment response, and that change in TBR will mediate response. Research Domain Criteria and EEG brain changes will be explored, including relationship of TBR to clinical symptoms, executive-function impairment, and sleep.